Post 1 — When the Metric Became the Mission

Every reward system faces the same risk: the metric can be satisfied without producing what the metric was meant to measure. When that gap opens, the behavior follows the reward — not the mission. The gap is almost never visible until after the fact.

Duolingo is a language-learning platform — forty million daily users, a market cap that once topped $7 billion. The company built its product entirely on behavioral rewards: streaks (daily login chains), XP (experience points), leaderboards, hearts. Log in, complete a lesson, keep your count alive — the app rewards showing up. Duolingo knew more about how rewards shape behavior than almost any company in operation. That story is not this one.

This one is about what happened when Duolingo applied that same logic inside the company. The reward drifted from the mission. The behavior followed exactly.

The streak is what made Duolingo famous. Log in daily, complete a lesson, keep your count alive—and the app rewards you for showing up. But Duolingo also sells a streak freeze. For a fee, a user can preserve their streak on days they don’t study. The streak survives. The learning doesn’t. Duolingo says the freeze is about learner flexibility—and they have research to back it. They also charge for it. Either way, the structure is the same: the freeze lets a user gain the reward—the streak—without doing the work the reward was meant to encourage. Duolingo built a mechanism that separates a user’s reward from a user’s goal.

Keep that structure in mind, because they built it again—inside the company.

In January 2024, Duolingo cut approximately 10% of its contractor workforce. The stated reason: AI could now handle the work. No full-time employees were affected.

Sixteen months later, CEO Luis von Ahn made the strategy explicit. On April 28, 2025, Duolingo’s LinkedIn page published an all-hands email from von Ahn declaring the company would be “AI-first.” He compared it to Duolingo’s bet on mobile in 2012—a decision that won them the 2013 iPhone App of the Year. Then the memo laid out five constraints to “help guide this shift”:

• We’ll gradually stop using contractors to do work that AI can handle
• AI use will be part of what we look for in hiring
• AI use will be part of what we evaluate in performance reviews
• Headcount will only be given if a team cannot automate more of their work
• Most functions will have specific initiatives to fundamentally change how they work

(Von Ahn’s exact words. “Headcount” means permission to bring on new staff—a team could only hire if it first demonstrated it had automated as much of its work as possible.)

Each of those lines tells a team leader what they will be measured on and rewarded for. The memo also acknowledged the tradeoff the company had accepted: it would “rather move with urgency and take occasional small hits on quality than move slowly and miss the moment.”

The reaction was swift. Users threatened to delete the app; the backlash was significant enough that von Ahn told the Financial Times he “did not expect” it.

By August 2025, he told the New York Times the memo “did not give enough context.” By April 2026, Duolingo had dropped the measure from performance reviews entirely. Von Ahn explained the reversal on a podcast: “It felt like rather than being held accountable for the actual outcome, we’re trying to just push something that in some cases did not fit.” And: “At the end, we backtracked, and we said, ‘No. Look, the most important thing in your performance is that you are doing whatever your job is as well as possible. A lot of times AI can help you with that. But if it can’t, I’m not going to force you to do that.’”

Von Ahn named the problem himself: the metric had replaced the mission.

The streak freeze lets a user satisfy the reward without doing the underlying work—and earn Duolingo revenue in the process. The April 2025 memo created the same structure internally: AI use in performance reviews as the measurable reward, disconnected from whether using AI actually improved the work. Von Ahn’s own reversal is the record. The reward and the mission had come apart.

The company’s shareholder letters lead with: “to develop the best education in the world and make it universally available.” When you set a reward, you get the behavior it asks for. Whether that behavior produces what the organization stands for is a different question.

Duolingo found out the hard way. The pattern is not unique to them. It shows up in companies, in teams, in policy, in the rules people set for themselves. Wherever a reward can be satisfied without producing what it was designed to produce, eventually it will be. The only question is when someone notices.

In your organization, what metric is your team skilled at hitting — while the outcome that metric was designed to represent goes unmeasured?

If you’re working through an incentive design challenge in your organization — or trying to understand why a strategy that looked right on paper isn’t producing the results you expected — I work with leadership teams on exactly these problems. You can reach me at wayne@waynerepich.com.

Behavior Follows Rewards. The pattern shows up wherever people are measured and rewarded.

Subscribe for free. Share it with someone who needs to read it.

—Wayne

Last time: A Haifa daycare, Northern Ireland’s renewable heating program, and Wells Fargo — three different industries, one pattern.

Next: Wednesday, June 3 — Bonus post: Egypt has produced the dominant players in squash for three generations. The system that made it happen was designed on purpose.

In development — Olympic athletes disqualified for losing; a corporation whose internal ranking system made employees compete against each other instead of the competition; a pro sports league that rewarded losing for four decades. More on the way.

Sources

Duolingo all-hands email (LinkedIn, April 28, 2025 — authored by Luis von Ahn, posted by Duolingo)

CNN Business, January 9, 2024 — “Duolingo lays off staff as language learning app shifts toward AI”

Financial Times, 2025 — von Ahn comment on user backlash (“did not expect it”)

Fortune, August 18, 2025 — “Duolingo CEO admits his controversial AI memo ‘did not give enough context’”

Fortune, April 13, 2026 — “I’m not going to force you”: Duolingo CEO backs off from evaluating employees on their AI usage (reporting von Ahn’s remarks on the Silicon Valley Girl podcast, April 2026)

Duolingo, Inc. Q4/FY 2023 Shareholder Letter (8-K, Exhibit 99.2), filed with the SEC, February 2024 — mission statement confirmed

Duolingo, Inc. Q4/FY 2024 Shareholder Letter (8-K, Exhibit 99.2), filed with the SEC, February 27, 2025 — mission statement; DAU 40.5M (Q4 2024)

Duolingo, Inc. Q1/FY 2025 Shareholder Letter (8-K, Exhibit 99.2), filed with the SEC, May 1, 2025 — mission statement confirmed

Duolingo blog, Osman Mansur (Product), January 31, 2022 — “The habit-building research behind your Duolingo streak” (blog.duolingo.com) — streak freeze rationale; UPenn/UCLA research citation; +0.38% DAU lift from two-freeze policy

BFR is written to be accessible, welcomed, and celebrated by every reader — not simplified, not elevated. Just clear.